Análise Discriminante como Preditiva de Dificuldades Financeiras em Empresas Brasileiras do Mercado Acionário

Abstract

Verificar a situação econômico-financeira de uma empresa pressupõe a análise conjunta de suas demonstrações contábeis, no intuito de prever as condições futuras de gerar resultados e de honrar seus compromissos. O presente artigo tem como objetivo estimar funções discriminantes para os grupos de empresas lucrativas, intermediárias e deficitárias, listadas na BM&FBOVESPA, no período de 2009 a 2011. Os procedimentos metodológicos utilizados caracterizam o estudo como pesquisa descritiva, documental e quantitativa. Os dados utilizados foram coletados do banco Economática®. As empresas pertencentes ao setor financeiro e de serviços não foram consideradas na análise e aquelas que não apresentaram os dados requeridos foram excluídas, resultando em uma amostra de 255 organizações. A análise dos dados foi efetuada com o uso do software SPSS®, tendo como variável de reunião os três grupos de empresas e como variáveis explicativas os indicadores econômico-financeiros de liquidez, rentabilidade e estrutura de capital. A análise dos dados permitiu identificar a existência de separação entre os grupos, apontando a Composição do Endividamento como a variável que melhor representa essa separação. Duas funções foram criadas. A primeira delas que segrega as empresas lucrativas das intermediárias, apresentou alta capacidade para demonstrar as diferenças entre os grupos. A segunda função discriminante representou apenas o poder residual para segregação entre as empresas intermediárias e as deficitárias.

Economic-Financial Evaluation of Brazilian Companies With Open Capital in the Period 2011 to 2018

“…For this reason, the option was made to focus only on its results. This level of accuracy was very close to the estimates of the discriminant function reported in Teixeira et al (2013), who had an accuracy level of 82.8%. Table 7 shows the estimates of the coefficients of Equation ( 38), as well as the degree of importance of each variable in this equation.…”

“…The gold price, volume trading exchange, average purchase and sale of the exchange rate and monthly deposit were used as indicators. Finally, they reported that ANN was able to adequately predict, with 95% of correct answers, the income of the sampled companies Teixeira et al (2013) aimed to use the DA method in the estimation of functions to discriminate companies between profitable, intermediary and non-profitable companies, between 2009 and 2011. Variables used were liquidity (general, current and dry), turnover asset, assets return, net margin, equity return, indebtedness, debt composition and fixed assets.…”

“…The proposed models were also carried out by Teixeira et al (2013), Chung et al (2008), Chen and Du (2009) and Aydın et al (2015), and, in the last model, only the variables from previous years were added as an analysis proposal.…”

“…The ANN structured in a contemporary way with Equation ( 38), presented a percentage of correct answers also above 83%. This form of modeling was applied by Teixeira et al (2013), using only the discriminant analysis and reached the level of accuracy of 82.8%, in a sample composed of 255 companies listed in the financial market. This result can be corroborated by Chen and Du (2009), in which they obtained an accuracy of 82.14% with the use of neural networks relating the companies" situation.…”

“…Among these studies, Chung et al (2008) and Chen and Du (2009), who built neural networks as a model for predicting the financial health situation; respectively, of New Zealand industrial companies and Taiwanese companies listed on the financial markets of those countries. Teixeira et al (2013) used the discriminant analysis method to classify the financial situation of 255 companies listed on BMF&Bovespa. Aydın (2015) evaluated the use of artificial neural networks (ANN) in the identification of financial crises in the Turkish economy, with seven macroeconomic and financial indicators, between the years 1990 and 2014.…”